Safety-critical AI platform tools
26 listings are filed under Safety-critical AI platform. Framework tags on these listings, by count: IEC 61508 (9), IEC 61511 (7), Machinery Regulation 2023/1230 (2), ISO 10218 (2). Most-tagged industries: Other safety-critical, Cross-industry, Process industries. 2 are open source under an OSI-approved license. A tag appears only when the cited public page names it. Combine this type with other filters.
-
AADvance Control System
Distributed, scalable safety instrumented system with fail-safe to TMR architectures. Official page: IEC 61508 certified (TÜV and Exida); TÜV-suitable for IEC 61508 up to SIL 3 and IEC 61511-1. SIL 1–3 in the same system. Not an AI-management platform.
-
ABB High Integrity I/O (S800 and Select I/O)
ABB SIL3 I/O for System 800xA High Integrity. Select I/O High Integrity is listed as certified for IEC 61508-SIL3 and compatible with AC 800M controllers using SM812. Page does not name IEC 61511. Not an AI-management platform.
-
ABB System 800xA High Integrity
ABB describes System 800xA High Integrity as a TÜV-certified SIL3 safety system and a complete SIS, deployable with System 800xA or independently. Page says it gives flexibility to meet IEC 61508 and IEC 61511 requirements. Not an AI-management platform.
-
Amazon Bedrock Guardrails
Configurable AWS safeguards for generative AI: content filters, denied topics, word lists, PII/sensitive-information filters, contextual grounding and Automated Reasoning checks. ApplyGuardrail can run those filters without invoking a foundation model.
-
Azure AI Content Safety
Azure service with text and image APIs that detect sexual content, violence, hate and self-harm, plus Prompt Shields, groundedness detection, protected-material scanning and a Task adherence API. F0 and S0 tiers.
-
Check Point AI Guardrails
Check Point AI Guardrails (Guard API at api.lakera.ai/v2/guard) screens LLM and agent traffic for prompt attacks, data leakage, content violations and off-policy tool use. Model-agnostic; 100+ languages. Available as SaaS or self-hosted.
-
Cisco AI Defense
Cisco AI Defense: algorithmic red teaming across 200+ threat subcategories, AI Runtime Protection guardrails, and network-layer enforcement across cloud, VPC and on-premises AI POD. Datasheet maps findings to NIST AI-RMF.
-
Cloudflare AI Security for Apps
Cloudflare WAF detections for LLM apps (formerly Firewall for AI; docs updated 8 September 2026): PII, unsafe-topic and prompt-injection scanning of JSON requests to cf-llm endpoints. Model-agnostic. Enterprise paid add-on; the WAF must be enabled.
-
CODESYS Virtual Safe Control SL
Virtual safety controller in a Linux container. Official page: certified according to IEC 61508 for SIL 3 applications, without IEC 61508 hardware certification of the host. TÜV SÜD. Diversified Encoding dual-channel. Not a trained-model SIL.
-
DeltaV SIS
Safety instrumented system used across process industries for ESD, burner management, fire and gas and pipeline. Official brochure: meets IEC 61508 and IEC 61511 requirements for physical separation; built for IEC 61511 compliance; TÜV and exida certified for SIL 3 without restriction.
-
Google Cloud Model Armor
Google Cloud service that screens LLM prompts and responses for responsible-AI safety categories, prompt injection and jailbreak, and sensitive data. Templates set confidence thresholds. Works on Google Cloud or other cloud providers.
-
Guardrails AI
Apache-2.0 Python framework (pip install guardrails-ai; Python 3.10–3.13) that runs Input/Output Guards of Hub validators on LLM traffic and can emit structured LLM output. Optional Flask server via guardrails start. Version 0.11.0.
-
HIMA HIMax
Fault-tolerant safety controller for process-industry SIS including fire-and-gas. Official fire-and-gas page: TÜV-certified HIMax suitable for use up to SIL 3 and certified in accordance with IEC 61511 up to SIL 3. Not an AI-management platform.
-
Honeywell Safety Manager SC
Third-generation QMR (2oo4D) safety platform for ESD/SIS applications in the Process Control industry. TÜV Rheinland certified up to SIL 3. Official standards list includes IEC 61508/61511. Not an AI-management platform.
-
INTEGRITY RTOS
Separation-kernel RTOS with integrated platforms for automotive, industrial safety, medical devices and avionics. Certifications named: ISO 26262 ASIL D, IEC 61508 SIL 3, FDA Class II and Class III medical device approval, and DO-178B/C DAL A for INTEGRITY-178. Not an AI model verifier.
-
KUKA iiQKA.OS2
Operating system and virtual robot controller for KUKA industrial robots. Official page: AI-capable, ready for ISO 10218:2025 and new machinery regulations; compliant with the machinery regulations (MACH REG). LINUX-based. Not a trained-model SIL.
-
Llama Guard 4
Meta Llama Guard 4 (12B): natively multimodal safety classifier pruned from Llama 4 Scout. Classifies LLM prompts and responses against MLCommons hazard categories S1–S14. Runs on a single GPU; documented as a drop-in for Llama Guard 3 8B/11B. First output line is safe or unsafe.
-
Llama Prompt Guard 2
Meta Llama Prompt Guard 2 BERT classifiers (86M and 22M) that label a prompt string benign or malicious for prompt injection and jailbreaks. Drop-in replacement for Prompt Guard. Designed for Llama 4 Scout/Maverick and Llama 3. Llama 4 Community License.
-
LYNX MOSA.ic.AI
Deterministic CPU and GPU runtime for AI at the mission-critical edge. Combines MOSA.ic with CoreSuite; not a training platform. The 8 April 2026 launch names a path to DO-178C, not a completed product certificate.
-
NVIDIA NeMo Guardrails
Open-source Python toolkit (pip nemoguardrails; Python 3.10–3.13) that adds programmable input, dialog, retrieval, execution and output rails to LLM applications. Library plus optional FastAPI server. Apache-2.0.
-
Palo Alto Prisma AIRS
Prisma AIRS AI Runtime Security: Network Intercept (inline firewall) and API Intercept (REST API/Python SDK) for prompt injection, jailbreak, memory poisoning and tool misuse. Network Intercept deploys on public cloud or on-premises microperimeter.
-
Portkey Guardrails
Guardrails on the Portkey AI Gateway: input and output checks with deny, log, fallback and retry. Failed checks can pass with HTTP 246 or kill the request with 446. Feature comparison names Portkey-Managed SaaS; new fully air-gapped deployments are no longer offered.
-
QNX OS for Safety
Safety-certified microkernel RTOS (QOS 8.0) certified by TÜV Rheinland to ISO 26262 ASIL D, IEC 61508 SIL 3, IEC 62304 Class C and ISO/SAE 21434. C/C++ toolchains classified TCL3/T3. FAQ names car, surgical robot, drone, power plant, machinery and control systems. Not an airborne runtime.
-
SIMATIC PCS 7 Safety
Process-industry safety integrated with SIMATIC PCS 7. SIMATIC S7 Safety Matrix (Engineering Tool and Viewer) is described as a holistic approach to functional safety in accordance with IEC 61511 across the safety lifecycle. Architectures certified up to SIL 3. Not an AI-management platform.
-
Universal Robots PolyScope
Software platform for Universal Robots cobots, from simple cells to AI-ready systems. Safety FAQ: PolyScope 5 and PolyScope X certified to EN ISO 10218-1:2011; targeting ISO 10218-1:2025 by January 2027 aligned with EU Machinery Regulation 2023/1230.
-
Yokogawa ProSafe-RS
TÜV SIL3 safety instrumented system described as a safety PLC for the process industries. Complied functional-safety standards listed: IEC 61508, IEC 61511-1 and IEC 62061. Used in over 3,500 projects. Not an AI-management platform.